COST

For more updates and info, visit our blog.

Taking subjective probability as a starting point, this WG will look at issues around the way these uncertainty concepts are operationalized and validated to ensure that they can represent a rational group assessment of uncertainty. This covers topics such as elicitation methods for preferences and probabilities, scoring rules, and probabilistic inversion to get distributions on model parameters without operational definitions. These will take account of the practical context of real world elicitation, as highlighted in the work of WG4, for example the incentives/disincentives on experts and the extent to which scoring rules can mitigate this.  In addition, we shall initially aim at:

• A critical analysis of semi-qualitative methods as risk matrix based methods that have become established in certain industries like aviation, cybersecurity,... aimed at providing improved methodologies.  
• Modeling adversarial uncertainties, or trying to forecast what adversaries will perform, eg in a security context.
• The use of structured expert judgment within multivariate extreme value models.

Focusing on foundational issues, WG3 will also aim providing support to other WGs as requested.

WG3 Documents - 16.04.15

onepageradversarial.pdf

onepagerdeep.pdf

onepagerfattails.pdf

onepagerqualitative.pdf

onepagervalidation.pdf